The best model for your team? You haven’t invented it yet. | Ai2’s Tim Dettmers
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A newsletter roundup covering AI coding and productivity topics: Ai2's Tim Dettmers explains how his team built the SERA model with minimal compute by synthesizing training data from private codebases; Andrej Karpathy's agentic Zettelkasten approach to self-maintaining knowledge bases; the 'Water Town' agent swarm data stack for data teams; Google Gemma 4 running offline on iPhone; why using Obsidian flat files as AI memory doesn't scale; managing mental fatigue from multi-agent workflows; and Google's new spam policy against back-button hijacking.
Table of contents
1. Building a self-writing knowledge base2. The agent swarm data stack3. Bringing offline AI to the iPhone4. The problem with using Obsidian as an AI database5. Curing the AI brain fry epidemic6. A new operating model for AI7. Google fights back against hijacked browsersSort: